Reinforcement learning consensus control for discrete-time multi-agent systems

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

1 Scopus citations

Abstract

In this paper, the consensus control of leader-follower multi-agent systems is investigated. To achieve the consensus of the discrete-time multi-agent systems, the data-driven iterative neighbor and target Q-learning algorithm is proposed. To implement the proposed method, the actor-critic architecture with neighbor and target networks are employed to approximate the Q-function and control signal. The reasonable reinforcement signal and cost function are chosen from the environment. This method is independent on the accurate system model where most practical systems are too complicated to build the accurate models. Finally, the simulation example is given to demonstrate the effectiveness of the proposed approach.

Original languageEnglish
Title of host publicationProceedings of the 38th Chinese Control Conference, CCC 2019
EditorsMinyue Fu, Jian Sun
PublisherIEEE Computer Society
Pages6178-6182
Number of pages5
ISBN (Electronic)9789881563972
DOIs
StatePublished - Jul 2019
Externally publishedYes
Event38th Chinese Control Conference, CCC 2019 - Guangzhou, China
Duration: 27 Jul 201930 Jul 2019

Publication series

NameChinese Control Conference, CCC
Volume2019-July
ISSN (Print)1934-1768
ISSN (Electronic)2161-2927

Conference

Conference38th Chinese Control Conference, CCC 2019
Country/TerritoryChina
CityGuangzhou
Period27/07/1930/07/19

Keywords

  • Actor-Critic Networks
  • Consensus
  • Neighbor Networks
  • Reinforcement Learning
  • Target Networks

Fingerprint

Dive into the research topics of 'Reinforcement learning consensus control for discrete-time multi-agent systems'. Together they form a unique fingerprint.

Cite this